Home Search Collections Journals About Contact us My IOPscience

Cohesion

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1931 Proc. Phys. Soc. 43 461

(http://iopscience.iop.org/0959-5309/43/5/301)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 129.174.21.5
The article was downloaded on 11/05/2013 at 12:07

Please note that terms and conditions apply.

 T H E PROCEEDINGS OF
T H E PHYSICAL SOCIETY
VOL. 43, PART 5 September I, 1931 No. 240

COHESION
BY J. E. LENNARD-JONES, T h e University, Bristol
A Lecture delivered before the Society on May I, 193 I.

5 I. INTRODUCTION: NEW IDEAS I N PHYSICS

H E R E are in nature, as in politics, two opposing forces. One of these aims

T at a peaceful consolidation and the other at a more active and probably

more spectacular disruptive process. I n nature it is cohesion between
atoms which tends to produce condensation and solidification, and temperature
which tends to produce dissociation, first of solids into liquids and then into isolated
molecules, then of molecules into atoms and finally into electrons and protons (as
in the hotter stars).
Temperature is a manifestation of kinetic energy and cohesion of potential
energy, and the interplay of these two forms of energy is responsible for many of
the observed physical properties of matter. A knowledge of the nature of cohesion
is thus a necessary step towards an understanding of these physical properties.
In this lecture I propose to review some of the progress which the new wave
mechanics has made possible in this direction and, by way of introduction, to refer
to some of the new conceptions which have been introduced during the last few
years.
First, there is the main idea of wave mechanics that it is impossible to follow
the electron in all its ways. Instead of supposing that the electron of an hydrogen
atom steadily pursues for all time a definite track or orbit, we are now content to
h o w the probability that it will do this or do that. We represent the electron by
continuous distributions-probability patterns. T o each such pattern there corre-
sponds a definite average energy and normally the electron stays in the pattern of
lowest energy. The density of the pattern at any point of space is a measure of the
probability that the electron shall be found there. T h e probabilities become in-
finitesimally small outside regions measuring one hundred-millionth of a centimetre,
SO that the electrons of an atom may be regarded as almost always within this small
PHYS.SOC. XLIII, 5 31
 462 J . -E. Lennard-Jones
region. This is the interpretation to be given to the “size” of an atom from the
new point of view”.
The mathematical theory does not give these density patterns directly. The
wave equation gives the possible energy levels and certain mathematical functions,
t) usually called wave functions and denoted by #, associated with them. T h e quantity
here denoted by t,b is a function of the three coordinates used to specify the position
of the electron in space. When # is a real quantity, its square is interpreted as a
measure of the density of the probability pattern, just as the square of the amplitude
(and not the amplitude itself) is a measure of the intensity of a sound or light wave.
When $ contains the square root of - I , the square of the amplitude of t,b, viz.
4 +
I t) 12, or the product of with its conjugate complex function, viz. $4, is taken
to be the appropriate measure of the probability of a given configuration.

§ 2. PAULI EXCLUSION PRINCIPLE

The hydrogen is particularly simple to work out because there is only one
electron. It is equally easy to work out the patterns for one electron alone in the
presence of a nucleus of charge Ze. The patterns are similar to those of hydrogen,
but are contracted in the ratio I to Z everywhere.
If electrons are added one at a time to a nucleus of charge Ze until the whole
structure is neutral and thus complete, we must inquire what happens to the electrons
after the first. There is nothing in the wave mechanics, as formulated at present,
to prevent all the electrons taking u p the same pattern. T h e density of the electrons
might be regarded as identical and superimposed, except for the electrostatic re-
pulsion between them which would tend to swell out the pattern again.
We know from experience, however, that all the electrons of an atom do not
assume the same pattern, and we have had to invoke an entirely new principle,
first enunciated by Pauli, and now generally referred to as the exclusion principle.
This principle asserts that there are never more than two electrons in the same
pattern; it goes further by asserting that it is not sufficient to describe patterns in
terms of the three spatial coordinates of an electron, but that a fourth coordinate
must be introduced. This fourth coordinate can take only two discrete values, SO
that the patterns we have described up to now are to have a further property added
to them which will double their number. I t is as though the patterns had colour
as well as shape. Actually it is usual to attribute a spin to the electron, and to
suppose that the axis of the spin can take up either of two quantized directions.
T h e exclusion principle then asserts that there is only one electron, described by
its four coordinates, in every pattern. I n building up atoms by the addition of one
electron at a time, the electrons arrange themselves in those patterns which respect
* At thls stage of the lecture, a number of diagrams of patterns of the excited states of the
hydrogen atom were shown. It was intended to reproduce these, but since the lecture, a number of
similar pictures have been published by H. E. White, Phys. Rev. 37, 1416(rg31), to which reference
may be made.
 Cohesion 463
the exclusion principle, and at the same time take up the configuration of least
energy. T h e application of this principle accounts in a very beautiful way for the
main features of atoms and molecules.

0 3. PRINCIPLE O F IDENTITY O F ELECTRONS

I n this addition of electrons to an atom, we have tacitly assumed that each

electron persists in one pattern. But so far as is known, electrons are identical,
and a system in which two electrons are interchanged is indistinguishable from its
original state. This principle of the identity of electrons has proved to be of great
importance. A method must be found of introducing this fact into the mathematics,
and this was first done about the same time by Dirac and Heisenberg.
If we suppose that there are N electrons in an atom and that the wave functions N
..
associated with the patterns which they occupy are u l , U%, . uN,these functions uN
including spin as well as space coordinates, then the wave function of the whole

In this determinant ztN (I) indicates that the four coordinates of electron I are t o ufl(I)
be introduced into the N t h function, and so on. Now if two electrons I and z
are interchanged, the effect on the determinant is to interchange rows I and 2,
and this has the effect of changing only the sign of Y. Furthermore, if we suppose
two wave functions u1 and u2 to be the same, two columns of the determinant
become identical and the determinant vanishes. T h e mathematical expression of
the Pauli exclusion principle is thus that no two of the wave functions ul,. . . ux
may be the same. T h e determinantal form of Y thus takes into account the identity
of the electrons and the exclusion principle at the same time.
Actually when the electrostatic repulsions between electrons in an atom are
taken into account, we cannot regard the wave function as made u p of N different
functions u1 . . . uN,each containing only the coordinates of one electron. All that
we can be certain of is that the complete wave function, when found, must have
the same properties as the determinant as regards the interchange of electrons and
the exclusion principle. It has not been possible, however, as yet to deal mathe-
matically with functions more complicated than the determinant.
The probability distribution of the N electrons is now given by YF,which

* This is subject to the condition that there be not several functions of this type with the same
energy. It is possible to write the wave function in the form given above for most atoms in their
State of lowest energy.
31-2
 464 J . E.Lennard-Jones
is a generalization of the probability distribution for one electron. When the two
determinants Y" and 'F are multiplied together, we find
YF =

p ( I , 2) where"

We note that p (I, I ) is the sum of terms like a9(I) uj (I) which are density functions
of electron I in the different patterns ; in fact, we have

and we should be justified in superimposing the density patterns if it were not for
the terms like p (I, 2). These indicate to us that we must no longer expect to express
the charge density in three dimensions (or four, including spin), but we must use
a six-dimensional function (or an eight-dimensional one if we include spin). This is
not surprising. I n the older theories, when electrons were assumed to be pursuing
orbits, we needed a knowledge not only of space coordinates but also of the com-
ponents of velocity of the electrons to specify their configuration completely. Now
that we have given u p the idea of attributing definite velocities and space coordinates
t o electrons simultaneously, we must not be surprised to find the six-dimensional
character of the probability configuration appearing in another form.

4. ELECTRON-DISTRIBUTIONS I N A T O M S
The expression given above for YT, multiplied by elements of volume
dr dTl, d ~. .~. d r, N , in the configuration space of each electron, may be interpreted
as the probability that an electron (we do not specify which one) shall be found
in d r l , another in dr2, and so on simultaneously-/-. T h e probability that an electron
shall be found in drl, irrespective of the position of the other electrons, is obtained
by integrating Y?F over the configuration space of electrons 2 to N. I n the inte-
gration each electron configuration is to be counted only once, so that if dr' in-
dicates one specified element of volume and dr" another, dr1' dr2" is to be taken to
be the same as dTl" dT2'. T h e result of the integration is p ( I , I) dT1 or Z p, (I , I) dT1,
3
and is thus the same as the superposition of the separate patterns p1 to p N f . I n this
* This form of the function qq has been given recently by Dirac, Proc. Camb. Phil. SOC.27,
240 ( 1 9 3 1 ) ; cf. also Lennard-Jones, Proc. Camb. Phil. Soc.27, 469 ( 1 9 3 1 ) .
. .t. P. A. M. Dirac, loc. cit.
$ This result holds when the functions ul to u.,. are normalized and orthogonal, that IS,

/%u,dr = Sa,. For most atoms in their normal state, the wave functions can be adjusted to satisfy
this condition.
 PHYSICAL SOCIETY, VOL. 43,PT. 5 (LENNARD-JONES)

Tofucep. 4651
 Cohesion 465
Sense we may construct pictures of many-electron atoms, and these pictures repre-
sent the probability of finding an electron in any element of volume in ordinary
space.
When the process is carried out, it appears that many atoms-more than were
previously suspected-may be considered to be spherically symmetrical. Of the
first ten atoms of the periodic table, six may be so regarded. They are hydrogen,
helium, lithium, beryllium, nitrogen” and neon. Diagrammatic pictures of some
of these atoms are given on the accompanying plate?. They have been calculated
from some approximate wave functions given by Slaterf. The pictures cannot be
regarded as accurate but they give a fair idea of the relative (‘spread ” of the various
atoms. T h e inert gas atoms, occurring at the end of a group of the periodic table,
are usually the most compact atoms of that group.
The spread of these atoms may be gauged also from figure I . T h e contours
give the probability of finding an electron outside the contour. T h e first is 0.2-
that is, the probability of finding an electron outside it is I in 5 . T h e second,
third. . . are 0.4, 0.6.. . . Outside the contour 1.0,there is a certainty of finding
one electron; outside the contour 1.2there is a probability of finding two electrons
simultaneously. T h e spread of lithium, which combines with others in the solid
state to form a metal, is in striking contrast to that of neon, although the latter
contains many more electrons. Boron in its normal state has one electron in a
2p-state, or one with quantum numbers n = 2, I = I, m = I , o or - I. T h e corre-
sponding probability patterns are asymmetrical. T h e one with quantum numbers
2 , I , o consists of two blobs of electron charge situated symmetrically with respect
to the nucleus. T h e plane of symmetry contains no charge, while a line through
the nucleus perpendicular to this plane of symmetry passes through the centres
of the blobs. Along this line the electron charge increases from zero to a maxi-
mum and then falls away to zeros, This line may be referred to as the axis of the
pattern. Those with quantum numbers (2, I , I ) and (2, I , - I ) are like anchor
rings. T h e concentration of charge in each is zero along the axis, and has a maxi-
mum along the core, from which it decreases to zero in each direction. T h e axis of
the ring coincides with the axis of the (2, I , 0)pattern so that the maximum con-
centration of the (2, I , I) pattern lies in the plane of zero concentration of the
( 2 , I, 0 ) pattern.
Carbon in its normal state may be regarded as having one of its two outer
electrons in the (2,I , 0)pattern and the other in either the (2, I , I ) or the (2,I , - I )
pattern, while the two electron spins are the samell. T h e electron distribution thus
consists of an anchor ring and two ((balls” of charge on its axis above and below
the nucleus, all of which is superimposed on a background due to the inner electrons,
which is spherically symmetrical.

* In its normal state, that is, in the 4Sstate.

I am greatly indebted to Mr H. H. M. Pike of the H. H. Wills Physical Laboratory, Bristol,
for these pictures, as also for those in figure I, and for the calculations involved in making them.
J. C. Slater, Phys. Rev. 36,57 (1930).
8 Cf. H. E. White, loc. cit. 11 The normal state of the carbon atom is a sPstate.
466 J . E . Lennard-Jones
Nitrogen has one outer electron in a (2, I , 0 ) pattern, another in a (2, I , I )
pattern and a third in a (2, I , - I ) pattern. These three patterns when superimposed
give a distribution which is spherically symmetrical. T h e spins of these three outer
electrons in the normal state of the nitrogen atom are the same, for this configuration
gives the lowest energy (the 4Sstate).

Fig. I. T h e “spread” of the atoms hydrogen, hehum, lithium and neon.

Neon, which has six electrons in the zp-state, has two electrons, one of each
spin, in each of the patterns (2, I , I ) , (2,I, 0) and (2, I , - I ) . It also is spherically
symmetrical. Flourine has one less electron than neon, and so we may regard it
as an atom which requires one electron to complete its spherical symmetry*. It
is an atom with a “hole” in it, and the “hole” is probably a (2, I , 0 ) pattern?.

* This property is connected with the affinity which fluorine has for an electron.
f Cf. J. E. Lennard-Jones, Trans. Faraday Soc. 25, 668 (1929).
 Cohesion 467
Oxygen has two " holes,'' one of which is probably a (2, I, 0)and the other a (2,I I),
or a (2,I , - I) pattern. Like divalent carbon, its normal state is a state, though
now this state may be attributed to two holes, whereas in carbon it is due to the
two electrons. I n this sense oxygen may be regarded as the counterpart of divalent
carbon.
These pictures of atoms throw considerable light on their chemical properties
and elucidate the problem of molecular structure, to which we shall refer below.

$ 5 . PRINCIPLE O F M I N I M U M ENERGY

To the new principles already discussed, we must add an old one, viz., the
principle of minimum energy; an atomic or molecular system tends to take u p the
state of lowest energy. This is of fundamental importance in the subject of cohesion.

$ 6 . TYPES O F COHESIVE FORCES

The physical and chemical properties of matter show that cohesive forces fall
into certain definite categories. Helium must be cooled down to - 269" C. before
it begins to aggregate, whereas hydrogen exists in the diatomic form at ordinary
temperatures. All the rare gases are similar to helium in exhibiting this weak
cohesion. Neon liquifies at about - 240' C. and argon at - 1 8 6 C ~ . T h e weak
attractive fields of these gases is responsible for their small departure from the
ideal gas laws, and as they were first investigated by van der Waals, they may
conveniently and fittingly be referred to as van der Waals attractive jields. This
type of attraction probably exists between all atoms and molecules, as will appear
in the next paragraph, but it is usually masked by other large attractive fields of a
different type.
The cohesive forces which hold together the atoms of a hydrogen or a nitrogen
molecule are about a thousand times as great as those which two helium atoms
exert on each other, even when they are neighbours in the solid form. These forces
are the familiar homopolar attractivejields of the chemist. At the same time it should
be pointed out that the molecules of hydrogen and nitrogen behave like the inert
gases as regards liquefaction and solidification and it is almost certain that such
molecules are held together in the condensed form by van der Waals forces.
It requires only about 500 calories to evaporate one gram molecule of solid hydro-
gen, whereas IOO,OOO calories per gram molecule are required to dissociate I gram
molecule of molecular hydrogen.
There is distinct evidence that in some solids, such as NaCl, there is a migration
of an electron from one atom (such as an alkali) to another atom (such as a halogen).
The atom which loses an electron becomes positively charged and the other
negatively charged. There is thus a net electrostatic attraction between neighbouring
atoms and this produces a very firm structure, which is difficult to disrupt. When
468 J . E . Lennard-Jones
such solids are vaporized (usually at high temperatures) the vapour is diatomic.
Such molecules may be said to be held by ionic attractivefields.
There are other solids which are also difficult to melt and to vaporize, but
which stand in a different class from the ionic solids or salts. They have other
properties not possessed by the salts in that they are good conductors of heat and
electricity and, so far as is known, are monatomic in the vapour state. These solids,
the-metals, stand in a class by themselves and the forces holding them together
may be called metallic attractive jields.
Intimately connected with the subject of cohesion is the question why atoms
ever repel each other. T h e very existence of matter leads us to postulate that when
atoms approach very near to one another they begin to repel. T h e resistance which
solids offer to compression is evidence and indeed a measure of the repulsive
forces between atoms. A theory which explains cohesion may be expected also
to explain intrinsic repulsive fields and this proves to be the case.

$ 7 . T H E N A T U R E O F VAN D E R WAALS F I E L D S
Although it is sufficient for many purposes to represent electrons in atoms as
continuous distributions, such representations are not, of course, accurate. It is
only the average which is continuous, and there must be rapid fluctuations about
this average, corresponding to the motion in a classical sense of the electrons
within the atom. Without it, van der Waals fields, as we know them, would not
exist. Two hydrogen atoms in their normal state, for example, are each represented
by a spherically symmetrical distribution, and as each is neutral, the electrostatic
potential of the one distribution on the other at a large distance apart is vanishingly
small. Nor can van der Waals fields be interpreted in terms of a static disturbance
of the continuous distribution. These fields have often been attributed vaguely
to polarization. But this polarization cannot be a static one, for if it were, each
atom so polarized could be represented by a small permanent dipole at its centre.
As the interaction in the case of equal atoms is symmetrical, the dipoles produced
must be symmetrical about a plane midway between them. They must therefore
point either towards or away from each other. I n either case they cause repulsion.
Van der Waals fields, as we shall show, are due to a dynamic polarization. The
motion of the electrons in one atom modifies that of the electrons in the other in
such a way that they tend on the average to move inphase. An important consequence
is that van der Waals fields t o a close approximation are additive, even when one
is surrounded symmetrically by several others. Although there was evidence from
many sources that these fields were additive, this property could not be deduced
from a theory of static polarization.
T h e nature of van der Waals fields may probably be brought out most clearly
by means of a simple molecular model. Suppose that an atom is represented by a
linear oscillator, that is, by an electron vibrating linearly about the nucleus. Such a
 Cohesion 469
system requires only one coordinate z to specify it. T h e appropriate Schrodinger z

where m is the mass of the electron, and &kz2is the potential energy of the electron.
On the classical theory, an oscillator of this type, subject as it is to a restoring force
of kz, would have a frequency given by
27rvo = vqz. VO
The solution of the wave equation is well known. It consists of a set of discrete
values for E, viz., E = ( n + 9)hv,, where n is any integer. I n the state of lowest n
energy, the appropriate value of $ is given by
+ = (2K voj7r)a e - K v o a p
where K~ = 27r mjh. K

The equation is written in such a form as to bring out the dependence of t,b on the
frequency vo for a purpose which will be seen later.
T h e “smeared-out” pattern for the electron in this case is thus a Gaussian
error curve, symmetrical about the origin.
We suppose now the oscillator to be subject to a uniform electric field F , so F
that the potential energy becomes *kz2 - eFz or 4k (z - eF/k)2 - e2F2/2k. T h e
appropriate wave equation can be written

which has the same form as the original equation, except that [, which = z - eF/k, E
now takes the place of z.
T h e energy values are E = (n + 4) hv, - e2F2/2k. T h e last term gives the
change of energy due to the field-the so-called polarization energy.
Polaarization energy = - e2F2/zk.
I n terms of the coefficient of polarizability cc, the polarization energy is - +aF2 a
by definition, so that cc = e2/k.
T h e corresponding density pattern is proportional to e-2rvo(s-eFik)a, which is the
same as the original one, except for a displacement of the maximum through an
amount eF/k in the direction of the field. This represents a static polarization.
We next suppose the oscillator to be subject to the field of a distant parallel
rigid dipole of strength p at a distance R. T h e potential of the electron is then
- 2 p z / R 3 and is like a parallel field except that the constant of proportionality
depends on R . T h e theory of the preceding paragraph applies if F is replaced by
2,u/R3, and so the energy of the oscillator, being proportional to the square of F,
is now proportional to V ~ R - ~ .
Polarization energy = - 2e2p2/kR6.
470 J . E . Lennard-Jones
Moreover, the energy being negative indicates attraction. T h e polarization of the
oscillator is again a static one.
If now we suppose the dipole p to fluctuate in value and to change in sign
independently of the original oscillator (though this is an artificial assumption), the
polarization of the oscillator is always such as to give an attraction proportional
to R-6; in fact, we have
Average polarization energy =- ze2p/kR6.
The polarization is now a fluctuating one and may be described as a dynamic one.
Actually the second dipole cannot be regarded as fluctuating independently
of the first. It also will be subject to a dynamic polarization, and the two systems must
be regarded as one complete coupled system. T h e method, however, prepares the
way for a more accurate treatment" and suggests that two such oscillators will
tend to move in phase.
T h e complete wave equation for two equal oscillators, each vibrating along the
axis of z under each other's influence, is

and this can easily be transformed to new variables

51 = ( z 1 + z,)/2/2, f2 = (21 - z,)/2/2,

where k, =k - ze2/R3, k, = k + ze2/R3.

T h e possible energy values of the equation are
E = (nl + 8) hvl + (v2+ 4)hvz,
where
v,
I
= - 2/(k,/m) = v,, $ / ( I
257
- ze2/kR3), v2 = v, Y'(I + 2e2/kR3).
This transformation is accurate and valid as long as v1 is real, that is as long as
2e2/kRS< I. This implies that R must be greater than (2a)B. For the state of
lowest energy, we thus get
E = +h (vi vg) +
= hv, ( I - e4/zk2R6),

if the square roots be expanded, as they can for large R . This indicates that the
energy of interaction of the dipoles, or the
van der Waals polarization energy = - hv,e4/zk2R6,
and that there is an attractive force proportional to the inverse seventh power of
the distance.
* Cf. F. London, Z e i t . f . phys. Chem. 11, 222 (1930).
 Cohesion 471
The wave function for the whole system for this lowest state is equal to the
product of the appropriate functions in f1 and f 2 , viz.,
9 = (4Ka v1v2,/Ta))a e - K ( Y 1 $ t P + I ' Z ~ B x )
- (zK/n)3(vlv2)t e-KVo (ziz+~az)e ( K v o e * / i k ) ( 2 ~ ~ 2 ~ i ~ ~ )

= (v1v2/v02)f e-(*vJk)v,

where #o is the corresponding wave function before interaction, and v is the potential
of interaction (- 2 8 2 z,z2/R3).
The probability of a specified configuration, that is, a specified z1and a specified
z 2 , is given by the square of +, and is thus proportional to $Jo2 e--8v. When v is
negative (indicating attraction between the dipoles), the probability is greater than
1cI02 and when v is positive (indicating repulsion between the dipoles), the probability

is less than $Jo2.

Fig. z . The probahility function for two interacting linear oscillators.

This result can be represented diagrammatically in a two-dimensional space, as

shown in figure 2. T h e distribution function, being proportional to e-2K (vlela+vaBaa),
is like a ridge with its maximum at the origin, and with elliptical contours. T h e
major axes of the ellipses are 5, = o or z1= + z 2 , and the minor axes z1= - z2.
The corresponding distribution without interaction is a round-topped mountain
with circular contours. T h e probability of finding z1and z2 with the same sign has
increased, while that of finding them with the opposite sign has decreased. I n other
words, the interacting dipoles tend on the average to move in phase.
A similar treatment can be given for two three-dimensional oscillators". Let
the line joining the centres of the oscillators be in each case the axis of z as before,
* F. London, loc. cit.
 472 J . E . Lennard-Jones
x, Y and let any convenient parallel sets of axes x andy be chosen in planes perpendicular
to this axis through the respective centres. The mutual potential energy of the
oscillators is then
+
V = e2 (x1x2 y1y2- zzlz,)/Ra.
The probability-distribution function can now be represented only in six-dimen-
sional space. It appears that z, and z2 tend to have the same sign (as before),
while x, tends to be opposite to x,, and y, opposite to y, . These are just the con-
figurations for which the oscillators attract. Hence we may say that two systems
tend to interact in such a way that the attraction is a maximum. It is in the genera-
lized sense that we use the expression “tend to move in phase,’’ for actually in
this example the x’s and y’s tend to be out of phase in the usual sense.
These simple examples, which can be worked out accurately, serve to bring
out the essential nature of van der Waals fields, and we may presume that in other
more complicated cases, where the mathematics is more difficult to interpret, the
attraction is due to a sympathetic movement of the electric charge.
The mathematical calculations for actual atoms are not easy. Wang* was the
first to attempt the calculation of the van der Waals fields of two hydrogen atoms.
He showed that the asymptotic form of the interaction of two hydrogen atoms was
like that of two dipoles, and used this additional potential as a perturbation of the
wave equation of two normal atoms. He thus showed that the attractive force
between them at large distances varied inversely as the seventh power of the dis-
tance. T h e numerical value has not, however, been confirmed by later investigations.
Recently Eisenschitz and London? have given a method of calculating fields of
this type for any atoms, which is based on the second approximation of the usual
perturbation theory. (The first approximation of this theory vanishes at large
interatomic distances). T h e wave function of the system is expressed in terms of
certain selected unperturbed wave functions of the two atoms. T h e method is
rather unwieldy and the authors themselves expressed the hope that some simpler
method would be found. T h e problem has also been considered by Lennard- Jonesf ,
HassC 4, and Slater and Kirkwood 11. T h e first of these authors uses a modified form
of the perturbation theory which considerably simplifies the calculation of the
van der Waals fields of two hydrogen atoms, while the second two use specially
adapted variation methods. I n each case the potential of the attractive field is of
the form - h R - 6 , and the values obtained for h are as follows :
Eisenschitz and London ... 6.04. I O - ~ O
Lennard-Jones ... ... 6.04. I O - ~ O
Hass6 ... ... ... 6-05. I O - ~ O
Slater and Kirkwood ... ... 6.05. I O - ~ O
These values of h give the attraction in ergs when R is measured in centimetres.
* S. C. Wang, Phys. Zeit. 28, 663 (1927).
t R. Eisenschitz and F. London, 2 e i t . f . Phys. 60,491 (1930).
1 J. E. Lennard-Jones, Proc. R.S. A, 129, 598 (1930).
5 H.R.Hass.5, Proc. Camb.Phzl. Soc. 27, 66 (1931).
/I J. C. Slater and Kirkwood, Phys. Reo. 37, 682 (1931).
 Cohesion 473
The calculations for hydrogen atoms are valid, of course, only at large inter-
atomic distances. At smaller distances the exchange phenomenon, referred to
below, comes into prominence, and leads eventually t o the formation of the diatomic
molecule. The exchange forces, however, fall off very rapidly (as e-ar/r) and
become small at distances at which the van der Waals fields are still important.
It is unfortunate that in this case, where the theoretical calculations can be carried
out with some accuracy, comparison with experiment is not possible.
An attempt has been made to extend the theory to more complicated atoms,
but the main difficulty here is the lack of knowledge of the wave functions of these
atoms. Even in the case of helium, the wave functions are only known approxi-
mately. London" has attempted to correlate the van der Waals fields of atoms with
their coefficient of polarizability, but the correlation is only an approximate one,
and only between certain rather wide limits can the magnitude of the fields be
fixed by this method.
HassCt has attempted the case of two helium atoms and has based his work
on wave functions deduced by Hyllerasf from a variation method. These wave
functions are adjusted to make the energy of the helium electronic system a mini-
mum and it does not follow that wave functions so deduced are valid over their
whole range; in fact, it appears that they are probably fairly accurate near the
nucleus (where the energy contributions are large) but less accurate in the outer
parts of the atom. But it is just the outer parts of the atom which are most affected
by outside fields, and so it is important that the wave function in those regions
should be known accurately. HassC finds thst the Hylleras functions, which
approximate best to the correct ionization potential of helium, do not give the
best r.esults for the polarization energy in a uniform electric field, which also is
known experimentally. Similar calculations have been carried out for the van der
Waals fields, but again there is a certain arbitrariness owing to the uncertainty
of the Hylleras functions. T h e results obtained by HassC are given below.
Slater and Kirkwoods have given a further development of HassC's method, also
depending on the variation method. They work out the polarizability of helium
and the van der Waals fields of two helium atoms by means of an approximate
wave function, previously given by Slater]].T h e results obtained so far for the
attractive-force-constant of helium (A in AI?-') are as follows : A
London ... ... ... ... 7 - 4 . 4 . 1 0 - 6 0 7
HassC ... ... ... ... (7*91-8*21)1 0 - 6 0
Slater and Kirkwood ... ... 8-94.I O - ~ O
Slater and Kirkwood give as well an expression for the field of two helium atoms
at close distances. T h e potential curve obtained from their work is plotted in figure 3.
* F. London, Zeit.f. phys. Chem.11, zzz(1g30).
t H. R. HassC, Proc. Camb. Phil. Soc. 27,66 (1931).
1 E. A. Hylleras, 2 e i t . f . Phys. 54, 347 (1929).
5 J. C. Slater and Kirkwood, Phys. Rev. 37, 682 (1931).
/I J. C. Slater, Phys. Rev. 32, 349 (1928).
T[ This figure is given as an upper limit, and is deduced from the formula given in the next
Paragraph, the observed value of the coefficient of polarizablllty being used.
 474 J . E. Lennard-Jones
Other inert gas atoms have been considered but the results can only be esti-
mated roughly because the wave functions are not known sufficiently well. Both
London and Slater-Kirkwood, unable to calculate the force-constants of these
gases directly, attempt to correlate the force-constant with the coefficient of polari-
zability. Slater and Kirkwood give
A = (const.) N* d,
while London gives A = (const.) I a2.

A, a: I n each case A is the constant of van der Waals field, and M. is the coefficient of
N polarizability; in the first formula N is the number of electrons in the outer shell,
I and in the second I is the ionization potential. It is not known yet how far either
formula is correct.
T h e attractive force constants of gases can also be determined from a study of
the equation of state, as has been shown by the author". A collected account of the
methods used and the results obtained has already been published?, but some new
results, which have been worked out recently on the assumption of an attractive
force of the type XR-', may appropriately be given heref. T h e repulsive field, which
A,rep ,, n comes into play at short distances, is represented by a force of the type A(rep.1 R-".
Theoretical calculations of the type discussed in the next paragraph show that the
repulsive field is more complicated than this and contains terms of the form e-uR,
but it falls off very rapidly with distance and can (in the case of helium at any rate)
be represented, over the range which is most effective in atomic collisions, by a
term of the type &rep.) R-".
T h e equation of state alone does not determine the value of the index n in
this repulsive field uniquely, but, for a given n, it determines the value bf the
A(att, constant Agep.), and the attractive constant A(att.1. Other methods have to be
employed to single out the right value of n from the array of possible ones thus
determined. T h e c r p t a l spacing of the solidified gas or its heat of sublimation
may be used for this purpose. I n the accompanying table the results are given
for n = IO, 11 and 13. T h e experimental material from which these results
are derived is cited elsewhere$, and later experimental results of Nijhoffll, and
of Gibby, Tanner and Masson7 have been used as well. I n table I are given
calculated values of the closest distance of approach which these gases would
have at absolute zero, if they set in the form of face-centred cubes. This is the form
. which would be expected theoretically for a force of the type Arep, R-n - Aatt.R-',
* J. E. Lennard-Jones, Proc. R.S.A,106,463 (19z4),107,157(1925),109,481(1925),112,214
(1926);J. E.Lennard-Jones and W. R. Cook, Proc. R.S. A,115, 334 (1927).
-f J. E. Lennard-Jones, chap. x of StatzstzcaZ.Mechanzcs, by R. H. Fowler (Camb. Univ. Press,
1929).
$ The author gratefully acknowledges the help of Miss M. J. Littleton (H. H. Wills, Physical
Laboratory, Bristol) in the numerical calculations.
§ J. E. Lennard-Jones, chap. x of Statistical Mschanzcs by R H. Fowler (Camb. Univ. Press,
1929).
I/ G. P. Nijhoff, Dzssert. (Leiden, 1928).
fi C. W. Gibby, C. C. Tanner and I. Masson, Proc. R S. A,122,283 (1929).
 Cohesion 475
Table I. Calculated force constants of gases from equation of state

calm-
lated 1 lated
heat Of Observ-
closest subh- Obser- ed heat
1
I
h(att.) distance mation ved of sub-
,
In crys- at ab!.
tal (A.U.)~zero in
lima-
spacing tion"
CalSJgm.

---
1 mol. I
n = IO 4.38.10-~~
Helium n = II 8.94. I O - ~ O
n = 13 4.55.10-105
__-
n = IO z-g4.10-~~
Neon n = II 6.285. IO-^^
n = 13 4*36.10-~O~
n = IO 1 7.19.10-80
Argon n = 11 1 2.10.10-87
1 n = 13 1 z.13.10-10?

Hydrogen 1 n = IO
IZ = 1 1
n = 13
1 4.59. IO-^^
I 16. IO-^^
7.79.10-104
1'07.IO+
3'33
0 * 8 5 . 1 0 - ~ ~ 3.28
0 . 6 3 . 1 0 - ~ ~ 3.18
~ _
450
478
529
_ -
_
-
-
_ _ _ ~
13.6.10-~~ 4:23 1380
II.I.Io-58 4.17 I460 4'0$ I860
8.38. IO-^^ 4.06 1640

as it is the cubic form of least potential energy§. Argon11 and neon7 have already
been showo experimentally to exist in this form. T h e calculated values of the heat
of sublimation at absolute zero from this crystal structure are also given. T h e
force constants of neon are a little uncertain because the experimental results
of Holborn and Otto and those of Cath and Kamerlingh Onnes are not very
consistent. T h e figures given are obtained by taking both sets of experimental
points and adjusting the theoretical curves to fit as well as possible. Figure 3
shows the curves of potential energy of pairs of inert gas atoms as a function of
their distance apart for the model n = 13, m = 7.
It is probable that many other gkses are held together in the liquid and solid
state by van der Waals forces, for, as London"" has pointed out, many cases are
known where the heat of sublimation is of the same order of magnitude as those
given in the above table. CH,, HC1, HBr, CO may be quoted as examples. Forces
of this type are probably responsible also in certain cases for the adsorption of

* For reference see F. London, 2 e z t . f . phys. Chem. 11,240 (1930).

t. J. de Smedt, W. H. Keesom and H. H. Mooy, Proc Amst. Acad. 3 3 , 255-257 (1930).
1 Vegard, Nature, 124,267, 337 (1929), molecules being regarded as spheres.
8 J. E.. Lennard-Jones and A. E. Ingham, Proc. R.S. A, 107,636 (1925).
11 F. Simon U. C. Simson, Zezt. f. Phys. 25, 160 (1924).
TI J. de Smedt, W. H. Keesom and H. H. Mooy, loc. cit.
** F. London, Zeit. f.phys. Chem. loc. cit.
476 J . E. Lennard-Jones
gases on solid surfaces. The author has shown that the order of magnitude of
certain observed heats of adsorption can be explained in this way*. Other cases
have been worked out recently by London?.

R (A.U.)

Fig. 3. The potential energy of pairs of inert gas atoms as a function of their distance apart in A.$
The experimental study of molecular spectra has established the existence of
molecules such as HgAr, HgKrg, and lately it has been shown that a molecular
form of K, exists quite other than the molecule formed by the usual homopolar
forces/].There is little doubt that these molecules are held together by van der
Waals forces.
* J. E. Lennard-Jones and B.M. Dent, Trans. Faraday Soc. 24, 92 (1928).
+ F. London, Zeit.f.phys. Chem. loc. cit.
1 The continuous curvesare obtained from the equation of state, hitherto unpublished, assuming
a law of force A(rey ,R-13 - R-'. The dotted curve for helium is obtained by Slater and Kirkwood
from its electronic structure.
5 0. Oldenburg, 2 e i t . f . Phys. 55, I (1929). \/ H. Kuhn, NuZurwissen. I S , 332 (1930).
 Cohesion 477
8 H O M O P O L A R COHESION’
In an earlier paragraph we have seen that when there are several electrons in
an atom, its configuration may in an approximate theory be specified by a number
of wave functions u l , u2 . ..
u N ,each a function of the four coordinates of an electron,
and that the probability of a specified configuration of the electrons is given by a
certain determinant of N rows and columns. T h e importance of the new form for
the probability function becomes evident when calculations are made of the energy
of electronic systems. T h e average value of the electrostatic energy of the electrons
can be calculated only when we know the probability that any pair of electrons
will be at a specified distance apart. A detailed calculation of the energy of an
atom or the interaction energy of two atoms depends then very closely on the
probability function w.
T h e probability of finding two electrons in specified places drl and dr2, in-
dependently of the position of the other electrons, is obtained by integrating the
probability function over the coordinates of all the electrons except two. T h e
result proves to be
b (I, 1) p (2,2) - p (I, 2) p (2, 1)) d.1 d72,
with the definitions of p (I, z), etc., given in $ 3 . T h e mutual potential of two
electrons consists then of two terms ;the first is the average of (e2/r12)
p (I, I) p (2,Z )
integrated over the whole of space of electrons I and z ; the second is the average
of (e2/r)p (I, 2) p (2,I). Apart from certain terms of these expressions which
cancel, the first represents the Coulomb interaction of the individual distributioiis
of electric charge u1iil, u2iiz, etc. This may be called the Coulomb energy. It is the
mutual electrostatic potential energy of the superimposed patterns which go to
make up atoms, as described in 5 4.
T h e second term in the above expression is new. It is difficult to describe its
physical nature. All that can be said of it is that it is the natural outcome of the
introduction of two new physical concepts into the mathematical scheme, viz.,
the principle of the identity of electrons, and the exclusion principle of Pauli. It is
sometimes described as the “ exchange ’)term and the term in the energy expression
arising from it as the “exchange” energy.
T h e ‘ I exchange” energy depends on the spin of the electrons, while the COU-
lomb energy does not. It is this property of the exchange term which has made it
of so much importance in the theory of atoms and molecules. I n a two-electron
system, as in an excited helium atom, for instance, the electrons may have the
same or opposite spins, and, owing to the exchange term, the energies of the states
with the same spins are lower in every case than those of the corresponding states
with opposite spin. T h e hydrogen molecule is another two-electron system, and
here the energy is lowest when the electrons have opposite spin. This appears to
’ The account of homopolar forces given here is necessarily brief. A fuller account has been
Wen recently by the author in a lecture to the London Mathematical Society on “The Quantum
Mechanics of Atoms and Molecules.” This is to be published in the Journal of the London Muthe-
matical Society shortly.
PHYS. Soc. XLIII, 5 32
478 J . E . Lennard-Jones
be the case in the interaction of most atoms. For certain interatomic distances the
energy is lowest, and, therefore, the cohesion greatest, when the electrons of one
atom are “paired” with those of the other atom. T h e energy of two interacting
nitrogen atoms is highest when the spins of the three outer electrons of the atoms
are all of them the same, and lowest when the three electrons of one atom are
paired with those of the other. This latter condition corresponds to the normal
nitrogen molecule, held together-as the chemist describes it-with a triple bond.
T h e pairing of electrons is thus brought into close connexion with the valency
rules of the chemist, and chemical homopolar forces are elucidated to this extent-
that they are seen to be a consequence of the same mathematical and physical
principles which have been formulated for other branches of physics. This result
may conceivably come to be regarded as one of the greatest achievements of the
present formulation of quantum mechanics.
T h e nature of these forces may be described in a simpler but less accurate way
by an appeal to Pauli’s exclusion principle. For one electron in the field of two
nuclei, there is a set of patterns with different energies just as there is for an electron
in the field of one nucleus. One electron in the presence of two hydrogen nuclei
will take u p the pattern of lowest energy. If a second electron is added to the system,
it also can take up the same pattern provided that it has a spin, which is opposite
from the first one. The two electrons now exist in the same molecular pattern and
are paired. Provided the energy of the pattern, which depends on the internuclear
distance, has a minimum, a stable molecule will be formed.
If, however, two helium atoms are brought together the electrons are paired
already. When the nuclei are pushed together, two of the electrons must be
“promoted” to a higher energy level if the exclusion principle is to be preserved.
T h e act of promotion requires energy and this appears as repulsion. A similar
argument may be used for other closed electron groups, which are brought into
contact.
T h e more accurate method described above also shows that closed electron
groups, like those of the inert gases, repel each other. T h e exchange term in this
case becomes predominant and leads to an increase of energy as the two systems
are brought near together. Intrinsic repulsive fields may be attributed to this new
exchange term, which quantum mechanics has produced.
T h e exchange term is more important than the Coulomb term in the interaction
of two hydrogen atoms, but the Coulomb term becomes relatively more and more
important in the interaction of larger atoms. This is due not only to the increased
spread of the atoms, but also to the form of the patterns which the outer electrons have
to assume. T h e effect is found even in the interaction of two lithium atoms”. When
the Coulomb part is important, the pictures of atoms given in an earlier paragraph
are useful in constructing models of molecules and helpful in understanding their
various shapes. We have, for instance, seen that an atom of fluorine, or indeed
any halogen gas, has a “hole” in it. It follows that the nucleus is less screened in

M. Delbruck, Ann. d. Phys. 5, 36 (1930).

 Cohesion 479
Some directions than others. If the hole in the F atom is like the ( 2 , I , 0)pattern, the
atom is roughly like a spherical ball of electric charge with a double conical hole
scooped out of it. It is then not difficult to understand why a hydrogen atom can
combine with it t o give a HF molecule. T h e hydrogen’s electron cloud sinks as far
as possible into the hole of the fluorine, for in this position it gets as near as possible
to the fluorine nucleus. Of course, it must be stressed that this only provides a
rough picture. I n an accurate treatment, the wave function of the whole system
must be considered.

$9. I O N I C C O H E S I O N
T h e conception of a halogen atom as an inert gas atom with a hole in it is useful
in many ways. It has an affinity for another electron, and when one has been secured
the electron cloud becomes spherically symmetrical. Ionic salts, like NaC1, may be
pictured as arrays of spherical distributions, each on the whole charged, either
positively or negatively, and so held together by electrostatic forces. This brings
us to the question of determining a criterion between homopolar cohesion and
ionic cohesion. When are we to suppose that atoms become ionized in the process
of binding? The criterion is one of energy. Suppose that in figure 4 the curve

e I ntermolaoular Dirtmce

Fig. 4. Potential energy curves of atoms and ions.

(I) refers to the energy of two neutral atoms at various distances apart and that
curve (2) refers to that of the corresponding ions. T h e difference between the
energies ( I ) and (2) at infinite distance is equal to the difference of the ionization
potential of one atom (say an alkali) and the electron affinity of the other (say a
halogen). The curve (2)is then represented by (a - b/R)because of the electrostatic
attraction.
At the point of intersection A of the curves, the work put into the system at
infinity has been recovered, and at interatomic distances less than this the ionic
form has the less energy, If, however, the energy required for ionization is large,
the interaction is of the type (3). Then the ions have to be brought t o very close
32-2
480 3.E. Lennard-Jones
distances before the ionic form has less energy than the atomic, but the close
distances are not possible because of the intrinsic repulsion or the Pauli principle
referred to in the preceding paragraph.
Table 28 shows how this condition is fulfilled in the case of some typical ionic
compounds. T h e distance R refers to the value of the abscissa where the ionic
curve cuts the axis ; this is approximately the same as the place where it cuts the
curve (I) when R is large.
Table 2.
Ionization potential
+ electron affinity (L.U.)
(volts)
KF 0.24 60
KCl 0.50 29
KBr 0.84
KI 1-23

It is likely that the terms ionic and homopolar refer to extreme cases, convenient
for classification, but rarely existing i n practice, actual cohesion being neither the
one nor the other but partaking of both. T h e wave function of a perturbed system
must be expressed in terms of all the wave functions of the unperturbed system
and if there are two such wave functions of nearly the same energy, the wave
function of the molecule is a linear function of both.

§ IO. M E T A L L I C COHESION
From the point of view of wave mechanics, a metal must be regarded as an
enormous molecule, with an enormous number of energy levels and electron
patterns. T h e problem is to explain why a collection of atoms in the form of a
metal has less energy than the isolated atoms of which it is composed. Is the cohesion
due to ordinary electrostatic forces or to the exchange phenomenon or both?
T h e problem has been considered recently in a qualitative way by J. C . Slater?,
who has shown that it is likely that atoms of one spin are surrounded by others of
opposite spint. I n those metals, which are of the body-centred-cubic type of
structure, each atom is at the centre of eight others of opposite electron spin.
In the bringing of the atoms together in this way the patterns of the individual
atoms have been made to overlap. An electron which formerly belonged to one
nucleus is now brought under the influence of several. This has the effect of
increasing the electrostatic attraction, and even in the absence of any other cohesive
force the metal would hold together. T h e metal is like a sea of electric fluid, in
which the nuclei float like buoys.
# This method of illustration is due to F. London, 2 e i t . f . Phys. 40, 475 (1928).
t J. C.Slater, Phys. Rev. 35, 509 (1930).
This is not true, of course, for the ferromagnetic bodies.
 Cohesion 481
Now we inquire how the exchange phenomenon affects the cohesion. It is not
difficult to show that the state of a crystal, in which all the electron spins are
perfectly paired, combines only with the states in which two of the spins are
interchanged. Thus, to use a linear example, the electron spins being denoted by
and p, the state ap . .
ap. . combines only with ap pa ap. , and similar arrays.
In the body-centred-cubic lattice the effect of such an interchange is to surround
two electrons with seven neighbours of the same spin. T h e resulting structure has
then considerably higher energy, because the contiguity of like spins causes a
repulsion. It is thus highly improbable. T h e electron seems to know what will be
the result of an interchange and decides that it simply is not done-or at any rate
only with moderation. T h e net result is that the exchange effect is not as important
in metals as it is in diatomic molecules, whereas the electrostatic attraction is relatively
more important. An accurate quantitative treatment of the problem is, however,
still required.
There are some metals, such as zinc and cadmium, in which the exchange term
probably makes little or no contribution to the cohesion. T h e atoms of which they
are composed have paired electron spins even when widely separated, and may be
expected on that account to repel like inert-gas atoms when brought into contact.
The fact that the metals are diamagnetic is further evidence that the metals
consist of perfectly paired electron spins and are formed from normal unexcited
atoms. Now the diatomic molecules Zn,, Cd, , are known to exist and the order
of magnitude of their heats of dissociation, which is known from molecular spectra,
suggests that these molecules are held together by van der Waals fields. T h e
cohesion of the corresponding metals is several times greater, but this is probably
due to the larger number of immediate neighbours and the additive property of
van der Waals fields. Thus it seems likely that van der Waals fields make a greater
contribution to the cohesion of some metals than has yet been realized.
Metallic cohesion cannot therefore be classified in any simple way. It is pro-
bably due partly to the electrostatic interaction of space-charge distributions,
partly to the exchange phenomenon, and partly to van der Waals fields. T h e re-
lative extents to which these various factors contribute to the cohesion of a metal
must vary from case to case and is still a matter for investigation.

§ 11. C O N C L U S I O N

I n conclusion we may summarize the contributions of the new quantum

mechanics to this branch of physics under four headings:
(I) It has provided density pictures of atoms which have permitted the
evaluation of the electrostatic interaction of atoms in a way quite impossible in an
orbital theory.
(2) It has led to the introduction of a new concept-an exchange force, which
is responsible for the homopolar bonds of the chemist.
482 J. E , Lennard-Jones
(3) I t has provided pictures of asymmetrical atoms like the halogens and thrown
light on the nature of electron affinity. This affinity is responsible for ionic structures
of the NaCl type.
(4) It has provided an explanation of van der Waals fields, which formerly
were not understood.
The general principles seem to be established. What is now required is a
mathematical technique capable of applying them to particular cases.